Naver Cloud Achieves 61 Papers Accepted at Global AI Conferences in the First Half of the Year

Synergies in technology among Naver AI organizations such as Clova and Papago, integrated around Naver Cloud, are beginning to take full effect.

On the 13th, Naver Cloud announced that 61 papers were accepted at the most prestigious global artificial intelligence (AI) conferences in the first half of this year. This pace matches last year's record of publishing 100 papers in a single year.

In particular, Naver Cloud is expanding its technological influence in the field of large-scale AI. Major research in the large-scale AI field has been selected not only for conference acceptance but also as 'Best Paper' and 'Top 25% Research.' Since presenting the research paper on Korea's first large-scale AI HyperClova at the world's top natural language processing conference 'EMNLP' 2021, a total of 20 research papers related to large-scale AI from Naver and Naver Cloud have been approved for publication at global conferences.

In the first half of this year, Naver Cloud presented meaningful research across various fields, from the development of large-scale AI technology ecosystems to service applications and research for creating social value. A study proposing an evaluation metric to measure the performance of image generation models was recognized as 'Top 25% Research' based on paper review scores at ICLR 2023, the world's leading machine learning conference. Through this research, Naver Cloud introduced a metric to measure the diversity and creativity of images generated by large-scale AI, which is expected to serve as a foundation for the development of more advanced large-scale image generation models.

As large-scale AI began to be actively utilized in business by being introduced into various services beyond research areas, research on AI model compression, which has become important, was also accepted at the global machine learning conference ICML 2023. A new quantization technique that can reduce the operating costs of large-scale language models was proposed, and it will be used to efficiently operate Naver's next-generation large-scale AI, HyperClovaX.

Leadership was also established in the field of AI ethics. A study proposing a training dataset and protocol to mitigate biased statements by large-scale AI on socially contentious issues was accepted at the global natural language processing conference 'ACL 2023.' This research was conducted over a year by experts from various academic fields including AI, social sciences, and law, who operated the 'AI Ethics Forum.' Additionally, an interview study demonstrating the potential for creating social value through Naver's AI check-in call service for elderly people living alone, 'Clova Care Call,' based on a large-scale language model chatbot service, won the 'Best Paper' award at CHI 2023, the top conference in human-computer interaction.

Yoon Sang-doo, head of Naver Cloud AI Lab, said, “We will continue cooperation with leading global research institutions such as Seoul National University, KAIST, and New York University to advance large-scale AI technologies including HyperClovaX, and through research collaborations with the University of T?bingen and the University of Washington, we will build more trustworthy large-scale AI.”

This content was produced with the assistance of AI translation services.